Welcome to the Dempsey Lab. We develop statistical methods for digital and mobile health, with the goal of helping researchers learn when, how, and for whom an intervention should be delivered. This page explains the lab’s research identity and the principles behind how we work. The rest of the handbook turns those principles into concrete project practices.
Many health interventions are not single decisions. They unfold over time: support may need to change as a person’s needs, context, or response changes. These settings produce intensive longitudinal data and raise difficult questions about experimental design, causal inference, and sequential decision-making.
Our research develops methods for questions such as:
We are a methods lab, but we do not develop methods in isolation. A strong project begins with a consequential scientific problem, defines the target of inference or decision clearly, develops the necessary theory and computation, and ends with evidence that another researcher can understand and reproduce.
Understanding before volume. We aim for a medium volume of careful, high-quality work. Running more analyses is not progress if we do not understand what they test or whether their output can be trusted.
Learn where the uncertainty is greatest. Research plans are hypotheses, not promises. We test the assumptions or technical steps most likely to change or stop a project early, when changing direction is still inexpensive.
Make the work legible. Code, mathematical formulations, decisions, and results should leave a trail that a collaborator—and your future self—can follow. Reproducibility is part of the research, not a cleanup step at the end.
Develop independence deliberately. New researchers receive substantial tactical guidance. Over time, each person learns to frame questions, choose the next informative step, diagnose problems, and drive a project. Independence is something we build together, not something you are expected to demonstrate on arrival.
Treat the lab as a team. We give candid, constructive feedback; share useful code and knowledge; and take one another’s questions seriously. A colleague’s success strengthens the whole group.
The lab is led by Walter Dempsey in the University of Michigan Department of Biostatistics. Our group includes doctoral students, postdoctoral researchers, alumni, co-advisors, and collaborators across statistics, data science, behavioral science, and health. Current members and their research interests are listed in the Lab Members section of the main site.
People may contribute to several projects in different roles. On each project, one person eventually serves as the Directly Responsible Individual (DRI): the person who maintains the clearest view of the project’s goal, current state, open questions, and next step. DRI is a project role, not a rank, and no one is expected to arrive already knowing how to do it. Lab Expectations: Your Role as a DRI describes how that responsibility develops.
Most research happens in project teams, supported by regular 1:1s, project meetings, and full-lab conversations. Meetings are most useful when the relevant working documents are current and the group can spend its time on interpretation, choices, and strategy rather than reconstructing what happened.
Each project therefore has a small set of shared homes:
The linked pages below explain how those pieces fit together. The organizing principle is simple: someone joining a project should be able to find its purpose, evidence, and next question without relying on one person’s memory.
Working hours. Research schedules are flexible, and no one is expected to clock in and out. At the same time, full-time students should treat the PhD program as a full-time commitment, plan for approximately 40 hours of work each week, be generally available during normal working hours (typically 9 a.m.–6 p.m., Monday–Friday), and attend scheduled meetings.
For an RA-funded PhD student, the usual allocation is approximately 20 hours per week for RA responsibilities and 20 hours per week for PhD research and training. The two may overlap intellectually, but they remain distinct responsibilities. Sustained progress is expected on both; being busy with one should not result in silently falling behind on the other. If the demands are in conflict, raise the issue early so that we can clarify priorities and adjust the plan.
Vacation. Students may take University holidays and up to 12 additional vacation days each year. Discuss vacation plans with Walter or your project supervisor well in advance and receive agreement before finalizing them, especially when travel overlaps with project deadlines, conference submissions, or responsibilities to collaborators.
Conference travel. Work-related conference travel is encouraged as part of research and professional development. Any request for registration, transportation, lodging, or other reimbursement must be discussed well in advance—before making reservations or financial commitments. We aim to distribute available travel support equitably and fairly across lab members. Support will depend on available funds and applicable University or sponsor rules, so reimbursement should not be assumed until it has been confirmed.
The DRI page describes project leadership. A smaller set of expectations applies to every lab member, regardless of role or seniority:
Problems are much easier to address while there are still options. Do not wait for a deadline, annual review, or small concern to become a large one before communicating. Raise concerns early and keep communicating as the situation changes.
This lab handbook complements rather than replaces the official student handbook and University policies. Consult your current student handbook whenever you have questions about program requirements, degree progress, funding or employment, time away, conduct, grievance procedures, or other student matters. If this site and an official policy differ, the official policy controls. Ask Walter, the program director, or the appropriate program staff early if you are unsure how a policy applies to your situation.
The handbook is a shared operating system, not a substitute for conversation. If a norm is unclear, outdated, or getting in the way of good research, raise it.